Express information identification method and device for express network, equipment and medium

Abstract

The invention discloses an express information identification method and device for an express network, equipment and a medium. The method comprises the following steps: acquiring an express network point image of a target express network point; inputting the express delivery network image into a pre-trained visual large model to obtain visual understanding information of which the output of the visual large model is the express delivery network image; inputting the visual understanding information of the express delivery network image into a pre-trained text large model to obtain predetermined scalar identification information of the express delivery network image, wherein the output of the text large model is the predetermined scalar identification information of the express delivery network image; and determining an express delivery information identification result of the target express delivery network according to the predetermined scalar identification information. Through combination of the visual large model and the text large model, an intelligent and efficient express information management solution is provided for a complex operation scene of an express network.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method, device, equipment and medium for identifying express information of an express delivery point. Background Art

[0002] With the continuous expansion of the logistics industry, especially the rapid growth of express delivery volume, the operating scenarios of express delivery outlets have become extremely complex. For example, each express delivery outlet is filled with various express parcels, including processed express parcels and unsorted express parcels, and express parcels are still being transported on the conveyor belt, etc. The existence of these situations makes it difficult for traditional image recognition methods to accurately identify the express delivery information of express delivery outlets.

[0003] However, in order to facilitate the unified management of all express delivery outlets, it is necessary to accurately identify the express delivery information of each express delivery outlet. Therefore, there is an urgent need for a solution that can perform intelligent and efficient identification in various complex scenarios of express delivery outlets. Summary of the invention

[0004] The present invention provides a method, device, equipment and medium for identifying express information of an express delivery network point, so as to solve the problem of inaccurate identification of express information in the express delivery network point facing complex scenarios.

[0005] According to one aspect of the present invention, a method for identifying express information of an express delivery outlet is provided, comprising:

[0006] Acquire the express delivery network image of the target express delivery network;

[0007] Inputting the express delivery outlet image into a pre-trained visual big model, and obtaining the output of the visual big model as visual understanding information of the express delivery outlet image;

[0008] Inputting the visual understanding information of the express delivery network point image into a pre-trained text model, and obtaining the output of the text model as predetermined scalar recognition information of the express delivery network point image;

[0009] The express information identification result of the target express outlet is determined according to the predetermined scalar identification information.

[0010] According to another aspect of the present invention, there is provided a courier information identification device for a courier outlet, comprising:

[0011] An express delivery outlet image acquisition module, used to acquire an express delivery outlet image of a target express delivery outlet;

[0012] A visual processing module, used for inputting the express delivery outlet image into a pre-trained visual large model, and obtaining the output of the visual large model as visual understanding information of the express delivery outlet image;

[0013] A text processing module, used for inputting the visual understanding information of the express delivery outlet image into a pre-trained text model, and obtaining the output of the text model as predetermined scalar identification information of the express delivery outlet image;

[0014] The express information identification module is used to determine the express information identification result of the target express network point according to the predetermined scalar identification information.

[0015] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0016] at least one processor; and

[0017] a memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the express information identification method of the express delivery outlet described in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the express information identification method of the express delivery outlet described in any embodiment of the present invention when executed.

[0020] The technical solution of the embodiment of the present invention provides an intelligent and efficient express information management solution for the complex operation scenarios of express outlets by combining the visual large model (VLM) and the text large model (LLM). The visual large model accurately identifies the visual data in the images of the express outlets, such as the area ratio of the piled goods, the dynamics of the items on the conveyor belt, the operating status of the staff, etc. The text large model performs reasoning and summarization based on the visual data to generate structured output. This combination of multimodal large models makes the express outlets' warehouse delivery process more intelligent and automated, ensuring that the packages can be processed and shipped out of the warehouse in a timely and efficient manner.

[0021] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 is a flow chart of a method for identifying express information of an express delivery outlet provided according to Embodiment 1 of the present invention;

[0024] Figure 2 is a flow chart of another express information identification method for an express outlet provided according to Embodiment 2 of the present invention;

[0025] Figure 3 is a schematic diagram of the structure of an express information identification device for an express delivery outlet provided according to Embodiment 3 of the present invention;

[0026] Figure 4 It is a structural schematic diagram of an electronic device for implementing the express information identification method of the express outlets according to the embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "candidate", "target", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] Embodiment 1

[0030] Figure 1A flowchart of a method for identifying express information at an express delivery outlet is provided for the first embodiment of the present invention. This embodiment can be applied to accurately identify express information at express delivery outlets in complex scenarios. The method can be executed by an express information identification device at an express delivery outlet. The express information identification device at the express delivery outlet can be implemented in the form of hardware and / or software. The express information identification device at the express delivery outlet can be configured in a server with communication and computing capabilities. Figure 1 As shown, the method includes:

[0031] S110: Acquire an express delivery network image of a target express delivery network.

[0032] Among them, the target express delivery outlets are the sites set up by the express company headquarters in various places to sort express deliveries in various regions. The express delivery outlet images refer to the scene image information of the express delivery outlets collected by the image acquisition equipment installed at the target express delivery outlets. Optionally, the express company headquarters can monitor the express delivery information in the target express delivery outlets by monitoring the express delivery outlet images, such as managing the express delivery sorting progress of the target express delivery outlets, or obtaining the working status of the couriers in the target express delivery outlets. However, due to the large number of target express delivery outlets corresponding to the express company headquarters, if the monitoring and management are directly carried out through the express delivery outlet images, the workload is huge. Therefore, it is necessary to identify the express delivery information from the express delivery outlet images. The express company headquarters improves the management efficiency of the target express delivery outlets by obtaining the identified express delivery information.

[0033] S120, inputting the express delivery point image into a pre-trained visual big model, and obtaining the output of the visual big model as visual understanding information of the express delivery point image.

[0034] Among them, the visual big model can understand the information in the image, accurately identify the important information in the express delivery point image, and describe the important information in text.

[0035] The visual big model is obtained by training with a pre-annotated dataset. Specifically, the standard COCO dataset warehouse is used as a pre-training sample set, which provides a large amount of general visual and language data for the model to help the model learn basic image features and language understanding capabilities. For specific network scene, multiple images are manually annotated. These annotated data contain specific object categories and scene information to provide targeted training data for the model. The COCO dataset is used for preliminary training to establish a basic visual big model. The goal of this stage is to enable the basic visual big model to recognize and understand various common visual features. After the preliminary training is completed, the basic visual big model is applied to the automatic annotation task. By using the automatic annotation function of the basic visual big model for unlabeled images, a large number (tens of thousands) of annotated samples are quickly generated. This method effectively expands the dataset and improves the robustness of the model. Determine the accuracy of the automatic annotation results of the basic visual big model to ensure that it meets the needs of the actual scene. Then, the basic visual big model is retrained according to the automatic annotation results to obtain a trained visual big model to improve the accuracy of the data.

[0036] Specifically, high-definition cameras and sensors are used to collect images and video data of express delivery outlets as training sample sets. Automated tools are used to clean and annotate the training sample sets, which improves processing efficiency. In addition, the training sample sets are expanded through data enhancement technology, which reduces the demand for large-scale sample sets and further reduces costs. After the training of the visual big model is completed, the express delivery outlet images of the target express delivery outlets are input into the visual big model to obtain the visual understanding information of the express delivery outlet images by the visual big model. The visual understanding information is the overall understanding of the express delivery outlet images. The visual big model identifies the image features in the express delivery outlet images through the feature learning results of the annotated information in the training sample set, and determines the visual understanding information based on the feature recognition results. The visual understanding information can be understood as an open result, which is a large paragraph of text understanding content generated by the visual big model based on the feature school results. For example, the express delivery outlet image includes several couriers, the working status of each courier, the express delivery information on the conveyor belt, the accumulation of unsorted express delivery, and the textual descriptive information of the sorted express delivery. The visual understanding information output by the big visual model is not conclusive information, but descriptive information about the image content.

[0037] The difficulty of the traditional model in accurately recognizing the images of express delivery outlets is that the scene complexity of the express delivery outlets causes the express delivery outlet images of each express delivery outlet to be not standardized, and the installation of the image acquisition equipment of the express delivery outlets is not uniform, resulting in the inconsistency of the corresponding content of the express delivery outlet images corresponding to different express delivery outlets. If the express delivery outlet images of all express delivery outlets are recognized by a single traditional model, it will cause the problem of inconsistent standards and inaccurate recognition of some express delivery outlets, and it is necessary to obtain the express delivery outlet images of all express delivery outlets for annotation during training, and it is difficult to collect sample data. For example, a transmission model is used to identify the image proportion of piled express parcels in the image, and then the express information recognition result is determined according to the image proportion. Among them, due to the inconsistency of the storage location of express parcels in the express delivery outlets, and they will be stored behind obstructed objects such as boxes and fences, it is difficult to recognize, which brings great interference to the image proportion result of the piled express parcels. Of course, there are also many interferences from complex problems in the scene that cause inaccurate recognition results. In this embodiment, the visual understanding information output by the visual large model belongs to the understanding information of the image content. The traditional visual model performs well in the laboratory, but once it is placed in a real express delivery outlet, the recognition accuracy will be greatly reduced. The scene environment of express delivery outlets is too complex: goods are piled up, conveyor belts are running at high speed, and people are walking back and forth. All these dynamic changes make it difficult for ordinary deep learning models to recognize accurate information. Among the piles of packages, the model must determine which packages have been processed and which have not yet been sorted, and even distinguish whether the personnel are performing the correct operations. This diverse and non-standardized scene makes it difficult for models to adapt and generalize, resulting in poor performance in practical applications.

[0038] In a feasible embodiment, the visual big model includes a lightweight visual big model and a heavyweight visual big model; the lightweight visual big model is obtained by distilling and training the heavyweight visual big model.

[0039] Since the overall framework of the visual big model is relatively large, the recognition efficiency of the visual big model for the express outlet images is reduced. Therefore, in order to improve the image recognition efficiency and improve resource utilization, when the information of the express outlet images is relatively simple, a lightweight visual big model is used to recognize the express outlet images. The lightweight visual big model is obtained by distillation training based on the trained complete visual big model. The distillation training can adopt any of the existing technologies, and the specific method of distillation training is not limited here. The complete visual big model is the heavyweight visual big model, that is, the visual big model with a complete framework.

[0040] S130, inputting the visual understanding information of the express delivery point image into a pre-trained text model, and obtaining the output of the text model as predetermined scalar recognition information of the express delivery point image.

[0041] Among them, the text big model can understand the text information, accurately identify the important information in the text information, and describe the important information through conclusive language. For example, the predetermined scalar recognition information output by the text big model is the result corresponding to the predefined recognition item. For example, the output result of the predefined text big model includes the quantity information of unsorted express delivery, the status information of the courier, the number of couriers, etc. Among them, the predetermined scalar recognition information is the recognition result corresponding to multiple predefined scalars, and the scalar refers to a fixed recognition object.

[0042] The text model is trained with a pre-annotated data set. Specifically, the output of the visual model is used as a pre-training sample set, which provides the model with a large amount of general visual and language data to help the model learn basic image features and language comprehension capabilities. The output of the visual model is annotated with predetermined scalar recognition information. These annotated data contain specific object categories and scene information to provide targeted training data for the model.

[0043] Specifically, the large text model learns the relationship features between the visual understanding information in the training sample set and the predetermined scalar recognition information, matches the relationship features of the visual understanding information of the express outlets images based on the learned relationship features, and summarizes the recognition results corresponding to the predetermined scalar from the visual understanding information based on the matching results as the predetermined scalar recognition information.

[0044] In a feasible embodiment, the text big model includes a lightweight text big model and a heavyweight text big model; the lightweight text big model is obtained by distilling and training the heavyweight text big model.

[0045] Since the overall framework of the text big model is relatively large, the recognition efficiency of the text big model for visual understanding information is reduced. Therefore, in order to improve the recognition efficiency and improve resource utilization, when the information of the express outlet image is relatively simple, a lightweight text big model is used to recognize the visual understanding information. The lightweight text big model is obtained by distillation training based on the completed text big model. The distillation training can adopt any of the existing technologies, and the specific method of distillation training is not limited here. The complete text big model is the heavyweight text big model, that is, the text big model with a complete framework.

[0046] In a feasible embodiment, S120 and S130 include:

[0047] Determine the current status parameters of the target express delivery outlets;

[0048] If the current state parameter is a peak state, the express network point image is input into a pre-trained lightweight visual large model, and the output of the lightweight visual large model is the visual understanding information of the express network point image; the visual understanding information of the express network point image is input into a pre-trained lightweight text large model, and the output of the lightweight text large model is the predetermined scalar recognition information of the express network point image;

[0049] If the current state parameter is a low-peak state, the express outlet image is input into a pre-trained heavyweight visual model, and the output of the heavyweight visual model is the visual understanding information of the express outlet image; the visual understanding information of the express outlet image is input into a pre-trained heavyweight text model, and the output of the heavyweight text model is the predetermined scalar recognition information of the express outlet image.

[0050] Among them, the current state parameter is used to characterize the current business busyness of the target express outlet. The current state parameter includes a peak state, indicating that the target express outlet is in a busy state; the current state parameter includes a low-peak state, indicating that the target express outlet is in an idle state. The current state adoption number can be determined according to the current time information. If it is determined according to the current time information that it is in the shopping promotion stage, it is determined to be in a peak state, otherwise the current state parameter is determined to be a low-peak state.

[0051] The embodiment of the present invention realizes intelligent selection of the most suitable model according to different warehouse delivery business scenarios of express delivery outlets. For example, in the case of rapid sorting during peak hours, the system gives priority to the high-efficiency model to ensure timely completion of warehouse delivery, and the high-efficiency model is a lightweight model; while in the case of low load, the high-accuracy model, that is, the heavyweight model, is selected to reasonably utilize resources and ensure that the warehouse delivery operation always flows efficiently.

[0052] In a feasible embodiment, determining the current status parameters of the target express delivery outlet includes:

[0053] Perform preliminary pre-recognition on the images of the express delivery outlets to obtain the pre-recognition information of the express delivery piles;

[0054] The current status parameters of the target express delivery outlet are determined based on the express delivery pile pre-identification information and the current time information.

[0055] In order to improve the accuracy of the current status parameters of the target express delivery outlets, the current actual situation of the target express delivery outlets and the current time information are combined for determination.

[0056] Specifically, image recognition technology is used to perform preliminary pre-recognition on the images of the express delivery outlets to obtain the express delivery pile pre-recognition results, wherein the image recognition technology can adopt traditional image recognition technology.

[0057] The first state parameter is determined according to the express accumulation pre-identification information, the second state parameter is determined according to the current time information, and the current state parameter of the target express outlet is determined according to the weighted result of the first state parameter and the second state parameter. The weights of the first state parameter and the second state parameter can be determined according to the actual scenario requirements. For example, a first relationship table mapping the express accumulation pre-identification information and the first state parameter, a second relationship table mapping the current time information and the second state parameter, and a third relationship table mapping the weighted result of the first state parameter and the second state parameter and the current state parameter are pre-established, and the current state parameter of the target express outlet is determined according to the express accumulation pre-identification information and the current time information based on the relationship table.

[0058] S140. Determine the express information identification result of the target express outlet according to the predetermined scalar identification information.

[0059] Determine the express information identification result of the target express outlet based on the predetermined rule and the predetermined scalar identification information. Since the predetermined scalar identification information is the identification result of a predefined fixed identification object, the corresponding identification result determination rule is determined according to the predefined fixed identification object, and the express information identification result of the target express outlet is determined based on the rule. The express information identification result is output in a structured format (such as JSON format) for use by other API interface calls.

[0060] The embodiments of the present invention can not only optimize the outlet delivery process, but also serve as a powerful tool for headquarters time efficiency management. By real-time monitoring of the accumulation situation, sorting status, and personnel operation efficiency of each outlet, the headquarters can clearly grasp the delivery progress of each outlet, make decisions quickly, and optimize resource allocation. According to the results of express information recognition, detailed time efficiency reports and data analysis can also be provided to help the headquarters better formulate strategies to ensure that each outlet can complete the delivery task on time and improve the overall logistics time efficiency.

[0061] The express information identification result may be predefined according to the identification requirements. In a feasible embodiment, the express information identification result includes at least one of the following: express accumulation information, express delivery information and express handling personnel information.

[0062] The technical solution of the embodiment of the present invention provides an intelligent and efficient express information management solution for the complex operation scenarios of express outlets by combining the visual large model (VLM) and the text large model (LLM). The visual large model accurately identifies the visual data in the images of the express outlets, such as the area ratio of the piled goods, the dynamics of the items on the conveyor belt, the operating status of the staff, etc. The text large model performs reasoning and summarization based on the visual data to generate structured output. This combination of multimodal large models makes the express outlets' warehouse delivery process more intelligent and automated, ensuring that the packages can be processed and shipped out of the warehouse in a timely and efficient manner.

[0063] Embodiment 2

[0064] Figure 2 This is a flow chart of a method for identifying express information at an express delivery point provided by Embodiment 2 of the present invention. This embodiment updates and trains the visual large model of the above embodiment to improve the efficiency of express information identification. Figure 2 As shown, the method includes:

[0065] S210: Acquire an express delivery network image of a target express delivery network.

[0066] S220: input the express delivery point image into a pre-trained visual big model, and obtain the output of the visual big model as visual understanding information of the express delivery point image.

[0067] S230, inputting the visual understanding information of the express delivery point image into a pre-trained text model, and obtaining the output of the text model as predetermined scalar recognition information of the express delivery point image.

[0068] S240, updating and training the visual big model according to the predetermined scalar recognition information of the express delivery point image and the express delivery point image to obtain an updated visual big model.

[0069] In order to improve the recognition efficiency of express delivery information, the training sample set is determined according to the output of the text big model, and the visual big model is updated and trained again.

[0070] Specifically, multiple predetermined scalar recognition information output by the text big model is obtained, and the predetermined scalar recognition information and the corresponding express outlet images are used together as a training sample set to train the above-mentioned visual big model to obtain an updated visual big model, so that the updated visual big model learns the predetermined scalar recognition information and the image features in the express outlet images.

[0071] For example, by using the few-sample learning technology during training, the updated visual big model can quickly adapt to new warehouse exit scenarios, reduce the need for large amounts of labeled data, and reduce training costs and time. Combined with contextual learning, the model can complete complex warehouse exit reasoning tasks based on a small number of examples, ensuring that regardless of the size of the outlet or the scenario, it can respond to and process warehouse exit needs in a timely manner.

[0072] During model training, the model is trained through performance evaluation, hyperparameter adjustment, and continuous learning. Performance evaluation refers to evaluating the performance of the model in the automatic labeling task and calculating indicators such as accuracy, recall, and F1 score. Analyze the performance of the model in different scenarios and categories to identify areas for improvement. Hyperparameter adjustment refers to adjusting the training parameters and model configuration based on the evaluation results to optimize the performance of the model. For example, adjust hyperparameters such as learning rate and batch size. Continuous learning refers to the continuous introduction of new data and new scenarios through incremental training or continuous learning to improve the adaptability and generalization ability of the model.

[0073] S250, inputting the express delivery point image into the updated visual big model, and obtaining the output of the updated visual big model as updated predetermined scalar recognition information.

[0074] The images of the express outlets are input into the retrained updated visual big model. The updated visual big model learns the image features of the express outlets images through the feature relationship between the image features learned during training and the predetermined scalar identification information, and outputs the updated predetermined scalar identification information based on the feature relationship.

[0075] Specifically, the express delivery network image here is an untrained express delivery network image, and may be a current real-time express delivery network image of the express delivery network.

[0076] S260, determining the courier information identification result of the target courier outlet according to the updated predetermined scalar identification information.

[0077] The express information identification result of the target express network point is determined based on the updated predetermined scalar identification information based on the predetermined rule. Since the updated predetermined scalar identification information is the identification result of a predefined fixed identification object, a corresponding identification result determination rule is determined based on the predefined fixed identification object, and the express information identification result of the target express network point is determined based on the rule.

[0078] The technical solution of the embodiment of the present invention updates and trains the visual big model through the output of the text big model, so that the express outlet image can be directly recognized through the updated visual big model to obtain the express information recognition result without going through the text big model, thereby improving the recognition efficiency of the express outlet image, making the express outlet delivery process more intelligent and automated, and ensuring that the packages can be processed and shipped out of the warehouse in a timely and efficient manner.

[0079] For example, deploy the trained large visual model to the production environment and integrate it into the actual application. Configure the relevant service interface and data processing flow to ensure that the model can process the input image data in real time. Set up a monitoring system to track the operation status and performance indicators of the model in real time. Record the processing time and response time to ensure the stable operation of the system. Establish a user feedback mechanism to collect and analyze user feedback in actual use in order to further optimize the model and monitoring system.

[0080] The engineering implementation plan of the monitoring system includes:

[0081] Implementation of the Reasoning Service

[0082] Edge computing device configuration:

[0083] Choose an edge cloud platform that supports self-built services and configure edge devices such as Intel NUC and Intel Flex170 that support inference tasks.

[0084] Install and configure inference frameworks (such as OpenVINO, Pytorch) and self-diagnosis tools (such as Edge Agent) to ensure that edge devices can monitor and report performance, computing load, temperature and other information in real time.

[0085] Configure the network speaker alarm function. When it is detected that the outlet fails to leave the warehouse in time, the system will issue a reminder through the network speaker on the edge device to ensure that the problem can be handled in time.

[0086] Cloud-based inference deployment:

[0087] Configure GPU instances on the cloud service platform in the headquarters data center, and select NVIDIA T4 and V100 deployment instances to support high-performance inference tasks.

[0088] Deploy inference services and use a self-built AI management platform. A single card supports more replica nodes and is equipped with self-checking tools to monitor service status, resource usage, and performance indicators in real time.

[0089] Unified management platform:

[0090] The self-built AI management platform uses Kubernetes for container orchestration and manages the deployment of edge devices and cloud instances. The Kubernetes cluster is used for load balancing and service discovery to ensure that inference requests can be reasonably distributed between the edge and the cloud.

[0091] Integrate self-test data from edge devices and the cloud into a unified monitoring platform, and upload self-test reports and performance data to a centralized monitoring system via an API for real-time analysis and troubleshooting.

[0092] Distributed Inference Architecture

[0093] Load Balancer Settings:

[0094] Deploy Nginx or HAProxy as a load balancer, configure a multi-layer load balancing strategy, and evenly distribute inference requests to multiple GPU card computing nodes. Set up health checks and traffic monitoring to optimize the load balancing effect.

[0095] Task Scheduling:

[0096] Use Apache Kafka or RabbitMQ as a message queue to manage the distribution of inference requests. Configure a task scheduler (such as Celery or Airflow) to coordinate task execution on distributed computing nodes.

[0097] During distributed reasoning, design strategies to ensure that each GPU card is evenly loaded. Use CUDA's torch.nn.DataParallel or torch.distributed to achieve load balancing and synchronization across GPUs.

[0098] Batch service implementation

[0099] Scheduled tasks and batch analysis

[0100] Scheduled task configuration:

[0101] Use Apache Airflow or cron job scheduler to set up scheduled tasks to automatically process image data at a fixed time every day. Configure tasks to read and pre-process image data and store it in the data warehouse.

[0102] Configure the data processing pipeline to ensure that the entire process from camera to storage and then to analysis is efficient and automated.

[0103] Business report generation:

[0104] Generate business reports using Python scripts or Jupyter Notebook. Combine data analysis tools (such as Pandas) and visualization tools (such as Matplotlib, Plotly) to generate detailed business reports showing key indicators and trends.

[0105] Realize automatic generation and distribution of reports, and set up automatic email notification or system reminder function after report generation.

[0106] Big Data Analysis

[0107] Data Integration:

[0108] Upload image data to the big data platform. Use Apache Flink for data processing and analysis, and set up data processing jobs for data cleaning, feature extraction, and conversion.

[0109] Configure the big data platform to support various data formats and storage requirements, and integrate with the data warehouse to provide flexible data access and analysis capabilities.

[0110] Analytical Model:

[0111] Design and deploy analytical models to discover potential business bottlenecks in data. Use machine learning libraries (such as Scikit-learn, XGBoost) to train models and integrate models into data analysis pipelines.

[0112] Configure model monitoring and evaluation tools to track the predictive performance and accuracy of your models and trigger automated updates and retraining processes when model performance degrades.

[0113] Implementation plan for automated operation and maintenance

[0114] Continuous monitoring and optimization

[0115] Monitoring tool configuration:

[0116] Deploy Prometheus for real-time performance monitoring and collect metrics such as GPU usage, response time, and throughput. Use Grafana to create dashboards to display key performance indicators in real time.

[0117] Configure alert rules to automatically notify operators when performance degrades or anomalies occur. Use Alertmanager or OpsGenie for alert notifications.

[0118] Automation Optimization:

[0119] Configure automatic triggers (such as Jenkins, GitLab CI / CD) to retrain and fine-tune the model. Set up automatic deployment and rollback strategies to ensure a smooth transition to the new model.

[0120] Implement automated performance optimization processes, such as dynamically adjusting resource allocation for inference tasks to optimize the inference speed and accuracy of the model.

[0121] Anomaly detection and troubleshooting

[0122] Anomaly Detection:

[0123] Integrate ELK Stack (Elasticsearch, Logstash, Kibana) for log management and anomaly detection. Collect and analyze log data in the reasoning process in real time and set up automatic alarm functions.

[0124] Configure log analysis rules to detect abnormal patterns and error conditions, automatically generate reports and provide resolution suggestions.

[0125] Troubleshooting:

[0126] Design fault response processes to automatically generate fault reports and notify operations personnel. Configure automatic repair functions, such as automatically restarting services, redeploying models, or rolling back to a stable version.

[0127] Implement fault recovery strategies, such as multi-region redundant backup and failover mechanisms, to improve system reliability and fault tolerance.

[0128] Embodiment 3

[0129] Figure 3 This is a schematic diagram of the structure of an express information identification device for an express delivery outlet provided in the third embodiment of the present invention. Figure 3 As shown, the device comprises:

[0130] The express delivery network image acquisition module 310 is used to acquire the express delivery network image of the target express delivery network;

[0131] A visual processing module 320, used for inputting the express delivery outlet image into a pre-trained visual large model, and obtaining the output of the visual large model as visual understanding information of the express delivery outlet image;

[0132] A text processing module 330, for inputting the visual understanding information of the express delivery outlet image into a pre-trained text model, and obtaining the output of the text model as predetermined scalar identification information of the express delivery outlet image;

[0133] The express information identification module 340 is used to determine the express information identification result of the target express network point according to the predetermined scalar identification information.

[0134] The technical solution of the embodiment of the present invention provides an intelligent and efficient express information management solution for the complex operation scenarios of express outlets by combining the visual large model (VLM) and the text large model (LLM). The visual large model accurately identifies the visual data in the images of the express outlets, such as the area ratio of the piled goods, the dynamics of the items on the conveyor belt, the operating status of the staff, etc. The text large model performs reasoning and summarization based on the visual data to generate structured output. This combination of multimodal large models makes the express outlets' warehouse delivery process more intelligent and automated, ensuring that the packages can be processed and shipped out of the warehouse in a timely and efficient manner.

[0135] Optionally, the device further comprises a visual large model updating module, which is used to:

[0136] After obtaining that the output of the textual large model is the predetermined scalar identification information of the express outlet image, the visual large model is updated and trained according to the predetermined scalar identification information of the express outlet image and the express outlet image to obtain an updated visual large model.

[0137] Optionally, the device further includes a second express information identification module, specifically configured to:

[0138] After acquiring the express network point image of the target express network point, the express network point image is input into the updated visual macro model, and the output of the updated visual macro model is updated predetermined scalar identification information;

[0139] The express information identification result of the target express outlet is determined according to the updated predetermined scalar identification information.

[0140] Optionally, the express information identification result includes at least one of the following: express accumulation information, express delivery information and express delivery processing personnel information.

[0141] Optionally, the visual big model includes a lightweight visual big model and a heavyweight visual big model; the lightweight visual big model is obtained by distilling and training the heavyweight visual big model;

[0142] The text big model includes a lightweight text big model and a heavyweight text big model; the lightweight text big model is obtained by distilling and training the heavyweight text big model.

[0143] Optional visual processing module and text processing module, specifically used for:

[0144] Determine the current status parameters of the target express delivery outlet;

[0145] If the current state parameter is a peak state, the express network point image is input into a pre-trained lightweight visual large model, and the output of the lightweight visual large model is the visual understanding information of the express network point image; the visual understanding information of the express network point image is input into a pre-trained lightweight text large model, and the output of the lightweight text large model is the predetermined scalar recognition information of the express network point image;

[0146] If the current state parameter is a low-peak state, the express outlet image is input into a pre-trained heavyweight visual model, and the output of the heavyweight visual model is the visual understanding information of the express outlet image; the visual understanding information of the express outlet image is input into a pre-trained heavyweight text model, and the output of the heavyweight text model is the predetermined scalar recognition information of the express outlet image.

[0147] Optionally, determining the current status parameters of the target express delivery outlet includes:

[0148] Performing preliminary pre-recognition on the express delivery outlet image to obtain express delivery pile pre-recognition information;

[0149] The current status parameters of the target express delivery network point are determined based on the express delivery accumulation pre-identification information and the current time information.

[0150] The express information identification device of the express network point provided by the embodiment of the present invention can execute the express information identification method of the express network point provided by any embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method.

[0151] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations and do not violate public order and good morals.

[0152] Example 4

[0153] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0154] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0155] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0156] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0157] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 11 executes the various methods and processes described above, such as the express information identification method of the express delivery outlet.

[0158] In some embodiments, the express information identification method of the express delivery outlet can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the express information identification method of the express delivery outlet described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the express information identification method of the express delivery outlet in any other appropriate manner (for example, by means of firmware).

[0159] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific reference products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0160] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0161] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0162] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0163] The systems and techniques described herein may be implemented in a computing system that includes a backend component (e.g., as a data server), or a computing system that includes a switch component (e.g., an application server), or a computing system that includes a frontend component (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, switch components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0164] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0165] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0166] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying express information of an express delivery outlet, characterized in that: The method includes: Acquire the express delivery network image of the target express delivery network; Inputting the express delivery outlet image into a pre-trained visual big model, and obtaining the output of the visual big model as visual understanding information of the express delivery outlet image; Inputting the visual understanding information of the express delivery network point image into a pre-trained text model, and obtaining the output of the text model as predetermined scalar recognition information of the express delivery network point image; The express information identification result of the target express outlet is determined according to the predetermined scalar identification information.

2. The method according to claim 1, characterized in that After obtaining that the output of the large text model is predetermined scalar identification information of the express delivery outlet image, the method further includes: The visual big model is updated and trained according to the predetermined scalar identification information of the express outlet image and the express outlet image to obtain an updated visual big model.

3. The method according to claim 2, characterized in that After acquiring the express delivery network image of the target express delivery network, the method further includes: Inputting the express outlet image into the updated visual big model, obtaining the output of the updated visual big model as updated predetermined scalar identification information; The express information identification result of the target express outlet is determined according to the updated predetermined scalar identification information.

4. The method according to claim 1, characterized in that The express information identification result includes at least one of the following: express accumulation information, express delivery information and express delivery processing personnel information.

5. The method according to claim 1, characterized in that in, The visual big model includes a lightweight visual big model and a heavyweight visual big model; the lightweight visual big model is obtained by distilling and training the heavyweight visual big model; The text big model includes a lightweight text big model and a heavyweight text big model; the lightweight text big model is obtained by distilling and training the heavyweight text big model.

6. The method according to claim 5, characterized in that Inputting the express delivery outlet image into a pre-trained visual big model, and obtaining the output of the visual big model as visual understanding information of the express delivery outlet image; Inputting the visual understanding information of the express delivery network image into a pre-trained text model, and obtaining the output of the text model as predetermined scalar recognition information of the express delivery network image, including: Determine the current status parameters of the target express delivery outlet; If the current state parameter is a peak state, the express network point image is input into a pre-trained lightweight visual large model, and the output of the lightweight visual large model is the visual understanding information of the express network point image; the visual understanding information of the express network point image is input into a pre-trained lightweight text large model, and the output of the lightweight text large model is the predetermined scalar recognition information of the express network point image; If the current state parameter is a low-peak state, the express outlet image is input into a pre-trained heavyweight visual model, and the output of the heavyweight visual model is the visual understanding information of the express outlet image; the visual understanding information of the express outlet image is input into a pre-trained heavyweight text model, and the output of the heavyweight text model is the predetermined scalar recognition information of the express outlet image.

7. The method according to claim 6, characterized in that Determining the current status parameters of the target express delivery outlet includes: Performing preliminary pre-recognition on the express delivery outlet image to obtain express delivery pile pre-recognition information; The current status parameters of the target express delivery network point are determined based on the express delivery accumulation pre-identification information and the current time information.

8. A courier information identification device for a courier outlet, characterized in that: The device includes: An express delivery outlet image acquisition module, used to acquire an express delivery outlet image of a target express delivery outlet; A visual processing module, used for inputting the express delivery outlet image into a pre-trained visual large model, and obtaining the output of the visual large model as visual understanding information of the express delivery outlet image; A text processing module, used for inputting the visual understanding information of the express delivery outlet image into a pre-trained text model, and obtaining the output of the text model as predetermined scalar identification information of the express delivery outlet image; The express information identification module is used to determine the express information identification result of the target express network point according to the predetermined scalar identification information.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the express information identification method of the express delivery outlet described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the express information identification method for an express delivery outlet according to any one of claims 1 to 7 when executed.

Images